generated: '2026-09-04'
method: probed
source: >-
  https://onewillow.com/agents.md (published rule) and observed response headers from
  https://onewillow.com/api/ucp/mcp on 2026-09-04
limit_count: 0
note: >-
  Willow states that its agent endpoint is rate limited but publishes no number, window or quota.
  What it does return on every call is a running cost signal — shopify-complexity-score and
  shopify-complexity-score-v2 — which an agent can use to pace itself even without a documented
  ceiling. A tools/list cost 370 (v2 37) and a tools/call of search_catalog cost 4940 (v2 494),
  so calls are priced by complexity rather than counted flat.
policy:
  statement: 'The MCP endpoint is rate-limited per IP. Back off on 429 responses.'
  source: https://onewillow.com/agents.md
  scope: per-IP
  window: not published
  limit: not published
  burst: not published
limits: []
response_headers:
  - name: shopify-complexity-score
    observed: true
    example: '4940'
    description: Cost of the request just served, in platform complexity units.
  - name: shopify-complexity-score-v2
    observed: true
    example: '494'
    description: The same cost on the v2 scale.
  - name: retry-after
    observed: false
    description: Not seen on any 2xx response; not provoked on a 429.
exhaustion:
  status_code: 429
  source: https://onewillow.com/agents.md
  observed: false
standard_headers:
  ietf_ratelimit_headers: false
  x_ratelimit_headers: false